Scale collaborates with NVIDIA to power the next generation of LLMs with NeMo SteerLM

Key Points:

  • Scale AI collaborates with NVIDIA to power the next-gen LLMs for generative AI.
  • Scale’s Data Engine produces more training data than anyone else, with two million hours expected in 2023.
  • SteerLM tailors language model responses to individual preferences, revolutionizing education and gaming.
  • The open-sourced dataset enables the research community to develop language models and drive innovation.
  • Enterprise software developers can leverage NVIDIA’s toolkits and models for a variety of industry applications.


Scale AI and NVIDIA have joined forces to revolutionize generative AI with their collaboration on LLMs (Large Language Models). By utilizing Scale’s top-notch training datasets and NVIDIA’s NeMo SteerLM, which allows dynamic steering of models during inference, developers can create applications for a wide range of enterprise use cases, from education to chat bots to gaming.


Scale’s Data Engine is the secret sauce behind their high-quality LLMs. It powers these advanced models with expert-generated data and thorough model test & evaluation. In fact, Scale produces more data than anyone else, with a jaw-dropping two million hours (or 228 years) of data expected in 2023. That’s enough data to make your head spin!


The collaboration between NVIDIA and Scale brings together the latest research in generative AI and high-quality datasets. The result? LLMs that can be aligned to follow instructions. SteerLM, a shining example of this collaboration, has the potential to revolutionize education, gaming, and more. Imagine language models tailored to individual learning preferences or non-player characters (NPCs) in games with customizable personalities. The possibilities are endless!


To support the advancement of generative AI, NVIDIA and Scale have even open sourced their dataset. This treasure trove contains 37k samples with various dimensions like helpfulness, correctness, complexity, and verbosity. By releasing this evaluation data to the wider research community, they’re enabling further innovation in language models. No need to start from scratch, enterprise software developers can now leverage NVIDIA’s open source toolkits and advanced models for a variety of industry applications.



Prompt Engineering Guides



©2024 The Horizon